China introduces rules for AI companion services and emotional interaction
China has introduced new regulations for AI companion services, requiring providers to address emotional risks, limit excessive use and strengthen user control over interaction data.
China’s Cyberspace Administration, together with four other government bodies, issued new regulations on 15 July 2026 governing AI companion services that provide ongoing emotional interactions through text, images, audio or video.
The rules apply to AI systems designed for sustained emotional engagement but exclude task-focused applications such as customer service tools, workplace assistants, education platforms and scientific research systems.
Under the new requirements, providers must introduce measures to identify signs of emotional distress and intervene in crisis situations. Platforms must also limit excessive use, provide users with control over their interaction data, and allow users to copy or delete chat histories. Providers are required to implement security measures, including encryption and access controls, and cannot share user interaction data with third parties without consent unless required by law.
The regulations also address potential dependency risks. Providers must prevent designs that encourage addictive use, introduce reminders after more than two hours of continuous interaction and provide interventions when systems detect signs of excessive reliance. Users must also receive reminders that AI systems are machines and not real human relationships.
Chinese researchers and legal experts highlighted concerns that some users, particularly young people and older adults, may develop strong emotional attachments to AI systems. Liu Xiaochun from the University of Chinese Academy of Social Sciences said reminders could help users distinguish virtual interactions from real-world relationships, while Wang Zhechen from Fudan University pointed to growing concerns about children’s use of AI companions.
Following the announcement, several Chinese AI platforms, including Doubao, Qwen and Yuanbao, announced the suspension of some AI-agent features. Experts said the regulations could encourage the development of safer approaches to human-AI interaction, combining technical safeguards with greater attention to social impacts.
